Structure‐Guided Design of Peptides as Tools to Probe the Protein–Protein Interaction between Cullin‐2 and Elongin BC Substrate Adaptor in Cullin RING E3 Ubiquitin Ligases

Abstract Cullin RING E3 ubiquitin ligases (CRLs) are large dynamic multi‐subunit complexes that control the fate of many proteins in cells. CRLs are attractive drug targets for the development of small‐molecule inhibitors and chemical inducers of protein degradation. Herein we describe a structure‐guided biophysical approach to probe the protein–protein interaction (PPI) between the Cullin‐2 scaffold protein and the adaptor subunits Elongin BC within the context of the von Hippel‐Lindau complex (CRL2VHL) using peptides. Two peptides were shown to bind at the targeted binding site on Elongin C, named the “EloC site”, with micromolar dissociation constants, providing a starting point for future optimization. Our results suggest ligandability of the EloC binding site to short linear peptides, unveiling the opportunity and challenges to develop small molecules that have the potential to target selectively the Cul2‐adaptor PPI within CRLs.

Te resaA.F .C ardote and Alessio Ciulli* [a] Cullin RING E3 ubiquitin ligases( CRLs) are large dynamic multisubunit complexes that control the fate of many proteins in cells. CRLs are attractive drug targets for the development of small-molecule inhibitors andc hemicali nducers of protein degradation. Herein we describe astructure-guided biophysical approacht op robe the protein-protein interaction (PPI) between the Cullin-2 scaffold protein and the adaptor subunits ElonginBCw ithin the context of the von Hippel-Lindauc omplex (CRL2 VHL )u sing peptides. Twop eptides were shown to bind at the targetedb inding site on Elongin C, named the "EloC site",w ith micromolar dissociation constants, providing a startingp oint for future optimization. Our resultss uggest ligandability of the EloC binding site to short linear peptides, unveiling the opportunity and challenges to develop small molecules that have the potential to target selectively the Cul2-adaptor PPI within CRLs.
Cullin RING E3 ubiquitin ligases (CRLs) are key machinerieso f the ubiquitin proteasome system as they are responsible for catalyzing the final step in the ubiquitination cascade, in which au biquitin molecule is transferred to the substrate. [1] CRLs, of which over 230 are estimated in human cells, are responsible for approximately 20 %o ft he ubiquitin-dependent protein turnover in cells, being implicated in an umber of cellular processes acrossd ifferent organisms. [2] Thes ignificant roles of CRLs in several biological processes and human diseasesh as rapidly emerged, in particulari nc ancer, where the genes encoding for E3 ligase subunits and their native substrates are often found as oncogenes or tumors uppressors. [3] Currently, much focus is directed towardt argeting E3 CRLs with small molecules, such as inhibitors, to block the ligase activity; [4] molecular glues, to redirect E3 CRL activity toward neo-substrates; [5] and bivalent PROTACs, to induce targeted protein degradation. [6,7] These chemical modalities motivate the growing interest in studying this class of enzymes.W hile E3 inhibitors, molecular glues and PROTACs have been widely reported, to our knowledge there are only af ew examples of small moleculesd eveloped to disrupti nter-subunit assembly within CRLs. [8,9] Our work focused on probing ap articularp rotein-protein interaction (PPI) in the CRL2 VHL ligase. The central scaffold of the VHL ligase is Cullin-2( Cul2), which recruits at the N-terminal domain the von Hippel-Lindaup rotein (pVHL) as substrate receptor, through an adaptor subunit constituted by ElonginB (EloB) andE longin C( EloC) and at the C-terminal domain the RING box protein, Rbx1 ( Figure 1B). We were interested in an epitope of the PPI between Cul2 and the receptor-adaptor trimeric subunit composed by pVHL, EloB and EloC (VBC).T his PPI has been described as comprising three main points of interaction. [10,11] The first crystal structure (Ref. [10]) comprised VBC boundt ot he first helicalb undle of Cul2 NTD ,w hereas a recent crystal structure (Ref. [11]) reported by us comprises the whole CRL2 VHL complex.T he latter work unveiled the importance of hydrophobic residues for the tight binding affinity observedb etween VBC and Cul2 (K d = 42 nm). In this work, we focused our attention on the contact surface between the N-terminal tail of Cul2 and EloC, whichwerefer to as EloC site.
Ta rgeting PPIs with small molecules provides many opportunities for basic biologya nd molecular therapeutics, but the physicochemical nature of thesei nterfaces turns the ability to modulate them into ag reat challenge. [12][13][14] Therefore, the identification and development of binding ligands to protein surfaces, whether direct or allosteric modulators of PPIs, remains ad ifficult and unsolved problem.F ortunately,m uch progress has been made in recent years in this direction. In particular, it is becominge vident that the development of drug-like PPI inhibitors, and small-molecule ligands to protein surfaces, can greatlyb enefit from the availability of ap eptidic ligand to that binding site, which could be from the naturali nteracting partner or from synthetic sources. [15] We hypothesized that Cul2-derived peptides bindingt ot he EloC site could provide valuable insight on how to target the Cul2-VBC PPI. Previous work in our research group has led to the development of potent small-molecule disruptors of the pVHL-HIF-1a PPI based on the structure of pVHL bound to a 19-mer parental peptide derived from HIF-1a. [16][17][18] It was therefore an attractive strategy to explore the potential to apply a similar approach to other non-HIF binding surfaces on VBC.
Such peptidic ligands could inform on the nature and details of key interactions essential to achievea ffinity at the targeted binding site. They could also provideu seful displacementt ools to ensure specificity of interaction of compound series in ligand development campaigns. Furthermore, peptides are interesting candidates as PPI modulators themselves, presenting an umber of advantages over non-peptidic small molecules: biocompatibility,a nd low toxicityt ot he organism;c hemical flexibility,s uch as the ability to adapt to large and often flexible surfaces;m odularity,t hus enlarging the structural diversity, enhancing selectivity and leadingt ohigh potency. [19] Considering the important role of the N-terminal tail of Cul2 in establishingt he interaction with VBC ( Figure 1), it was hypothesized that short peptides able to reproduce this tail could recapitulate the interaction, providing tools to develop chemicalp robes and target this PPI. Based on the structural analysiso ft he Cul2-VBCi nterface, we first aimed to recapitulate the interaction using N-terminal Cul2 peptides varying from three to eleven residues. The peptides were synthesized and seven of the nine peptides were tested for binding to VBC by Biolayer Interferometry (BLI) (the 9-mer and1 1-mer peptides could not be tested because they formed aw hite precipitate in the conditions of the experiment). The results showed that at least six amino acids were required to observe binding to VBC. The binding event was recapitulated with the 6-,7 -, 8and 10-merp eptides;n evertheless, the binding affinities determined were quite weak (K d in the millimolar range, Figure 2). Thus, the next step was to enhance the binding affinity of the peptides toward VBC.
Considering the 8-mer peptide( MSLKPRVV) had the best fitting, tighter binding affinity and reproducibility in the previous BLI assay,i tw as chosen as template for an alanine scanningt o identify hotspots in the 8-mer peptide. By replacing, one at a time, all the amino acid residues by an alanine residuew eo bserved that upon replacement of Pro5 or Arg6, the binding of the peptidet oward VBC was totally lost ( Figure 2). This was in agreement with the structurala nalysis of the Cul2-VBC complex, [10] which suggestst hat Pro5 is responsible for the folding of the N-terminal tail of Cul2 upon itself and Arg6 is responsible for keeping this conformational arrangement by establishing intramolecular hydrogen bonds with the carbonyl groups from the backbone of Ser2 and Lys4.R eplacing Lys4 with alanine resultedi na10-foldd ecrease in the binding affinity but the replacement of other amino acids in the peptidew ith alanine did not seem to perturb the interaction asm uch. The alanine scanning results also implied thatt he leucine residue, of which the side chain inserts into the EloC site could be replaced without significant loss of binding affinity.
From the resultso ft he initial screen we learned that:1 )the preferential size of the peptide comprised eighta mino acids; 2) Pro5 and Arg6 werec ritical to assure bindingt oV BC;a nd 3) Leu3 could be replaced without major loss of binding affinity.T hus, we employed as tructure-based approacht oe nhance the affinity of the 8-mer peptide, by replacing the leucine with amino acids presenting bulkier side chain groups.As malll ibrary of 8-mer peptides containing as et of natural andn onnaturala mino acids replacing Leu3 was designed and tested for binding to VBC. The criteria for choosing these amino acids was to modulate the bulkiness of the side chain group, for example,w ec hose phenylalanine and tert-butylglycine, among  others, to increase the volumeo ccupied by the side chain. As before,t he peptides were initially tested by BLI forb inding to VBC ( Table 1). The BLI resultss howed that the EloC pocket could accommodate all the derivative peptides, except when leucine was replaced by tryptophan, which was probably overly bulky and was found to decrease the binding affinity relative to the parental peptide.I na ddition, it was also observed that the replacement of leucine with tert-butylglycine and dimethylcysteine (peptides G and J,r espectively) led to the dissociation constant breaking into the micromolar range. Particularly for peptide J,t he K d value determined by BLI was 0.3 AE 0.1 mm,w hich represents a6 -fold improvement in regards to the parental peptide A.T hese results encouraged fur-ther in-depthc haracterization of peptide J,w ith the aim to better understand its binding interaction.
The binding of peptide J to VBC was characterizedf urther by other biophysical techniques in addition to BLI, namely isothermalt itration calorimetry (ITC), AlphaLISAc ompetition assay,a nd protein-observed nuclear magnetic resonance (NMR), which all showedc onsistent results. Titration of peptide J into VBC by ITC resulted in the determinationo fK d = 5.28 10 À4 AE 0.65 10 À4 m ( Figure 3B). This three-digit micromolar K d corroborated the K d value obtained by BLI ( Figure 3A). In the AlphaLISAa ssay,p eptides A and J were used to disrupt the native interaction between VBC and Cul2. Both peptides A and J were found to displace Cul2 ( Figure 3C).
The IC 50 determined forp eptide J in the AlphaLISA displacement assay was 0.22 mm,avalue similar to the K d values for direct binding measured by ITC and BLI.H owever,t he AlphaLI-SA assay could not distinguish the binding of peptide J from peptide A,which had shown 10-foldw eaker binding affinity by BLI, presumably due to differences between thet wo assays. Nonetheless, the AlphaLISA resultsc learly validated the binding to the Cul2 binding site. Finally,w ep erformed chemical shift perturbation (CSP) experimentsb yp rotein-observed NMR and the resultss uggested that the peptide was bindingt o VBC. Additionally,t he data was in agreementw ith the AlphaLI-SA, suggesting that peptide J was binding to the EloC pocket ( Figure 4). Upon binding to ac ertain area of the protein, the peptidec hanges the chemical environment of the residues that surround it. These changes in the chemical environment are registered as peak shiftingo rdisappearance. The residues affectedb yt he binding of the peptidew ere identified based on the peak assignment available for VBC (provided by Dr. Mark Bycroft, Cambridge). [20] It was remarkable that whilst some peaks wereu ndeniably affected, others remained constant. Mapping the disturbed residues on the structure suggestedt hat the residues more affectedb yt he presence of the peptide were near the EloC pocket ( Figure 4B). There were also some other peaks shifting that correspond to residues in different areas of the protein. It is expected that amino acids at ac ertain distance from the binding site might rearrange upon binding of aligand and, hence,promote shifts.
The biophysical characterization suggestedm oderate binding to VBC and it disclosed the opportunity to develop these Cul2 peptides into high-affinity binders. The strength of the interaction was boosted about 4-fold simply by increasing the volumeo ft he side-chain fitting the EloC site. It is anticipated that the structure of Cul2 N-terminal tail bound to VBC can differ significantly when in the context of an 8-mer peptideo r in the context of the full-length protein.T he structural knowledge of the binding mode of peptide J in complex with VBC would help to inform molecular designt oe nhance binding. Considering this, efforts were taken into co-crystallizing peptide J with VBC, however,t od ate this has not been achieved.
In conclusion, it has proven challenging to targett his particular PPI using peptides. TheV BC-Cul2 interaction appearst o be at ertiaryP PI, [21] involving multiple epitopes.I nf act, as observedi nt he crystal structure, the interaction is directed by three points of contactb etween pVHL, Cul2 and EloC. Despite the N-terminal tail of Cul2 being important for the specificity of the interaction, by itself it could not drive at ight binding event. Therefore, peptides that target only one of these interaction sites would likely not be able to mimic the native inter-action and thus would not block the PPI site effectively.O ther drug discoveryt ools such as peptides tapling, [22] or tethering, [12] for example, might be helpful towardt his goal. Another approacht ot arget this kind of PPIs could be the use of bicy-  clic peptidesa st hey cover al arge surfacea rea and are able to closely mimic PPI features. [23] Our work demonstrates how demanding it can be to target extendedP PI regions, as opposedt ow ell-defined bindingsites. In this particularc ase, it suggestst hat the narrow and merely hydrophobic nature of the EloC site make it ac hallenging target site. The gained knowledge and tools will nevertheless inform future development of small molecules that could target this specific PPIa nd could be used as chemical probes to study Cul2-dependent CRLs.

Experimental Section
Protein expression:V BC ternary complex was described previously. [17] BL21(DE3) E. coli cells were co-transformed with the plasmid for expression of pVHL/SOCS2 and the bi-cistronic pDUET plasmid for expression of EloBC. As ingle colony of transformant was used to inoculate LB media for bacterial culture. Protein expression was induced with 0.3 mm IPTG (when OD 600 reached 0.8) at 24 8Cf or 18 h. Co-expression of these proteins resulted in the formation of the respective trimeric complex (VBC) that was then purified by two steps of affinityc hromatography,f ollowed by ion-exchange chromatography and finally by size-exclusion chromatography.T he His-tag was cleaved between the two affinityc hromatography steps with TEV protease. Following this protocol the yield of protein was about 15-20 mg per liter of culture. For the expression of 2 H, 15 N-VBC, the LB media was replaced with E. coli-OD2 enriched media (Silantes) and the yield dropped to 4mgL À1 . [24] Peptide synthesis and purification:the peptides were synthesized in an INTAVIS RespepSL automated peptide synthesizer using solid phase Fmoc chemistry.T he peptides were cleaved from the resin using as olution of TFA, water and triisopropylsilane (TIS) (92.5:2.5:5). The peptides were obtained as C-terminal amides and were purified by HPLC in basic conditions (0.1 %NH 4 OH) in agradient of 0-100 %o fa cetonitrile in water over 15 minutes. The purity and identity of the peptides was determined by LC-MS.
Biolayer interferometry:B LI experiments were performed in an Octet RED384 (FòrteBio). Biotinylated VBC (25 mgmL À1 )w as immobilized on Super Streptavidin-coated biosensor (FòrteBio). The experiments were conducted at 25 8C, in 20 mm HEPES pH 7.6, 100 mm NaCl, 1mm DTT and 0.02 %( v/v)T ween-20 buffer.T he response of the reference tips was subtracted from the signal to account for unspecific binding. The data points were fitted using a 1:1model.
AlphaLISA:A nti-His 6 acceptor beads and Streptavidin donor beads (PerkinElmer) were used. The competition assay was performed in a3 84-well plate by mixing V 6 His BC (500 nm)a nd biotinylated Rbx1-Cul2 (150 nm)a nd titrating the competitor (peptide). The final volume of each well was 20 mL. The plate was then read in a PHERAstar FS (BMG LABTECH). Each of the competitors was titrated in quadruplicate. The fitting and IC 50 determination were performed in GraphPad Prism 7( GraphPad Software, La Jolla, CA, USA).
NMR spectroscopy:N MR experiments were carried out in an AV-500 MHz Bruker spectrometer equipped with a5mm CTPXI 1 H-13 C/ 15 N/D Z-GRD cryoprobe. The total volume of the sample was 200 mLa nd the experiments were performed in ac apillary tube containing 100 mm 2 H, 15 N-VBC samples in ab uffer of 20 mm KH 2 PO 4 pH 7.0, 50 mm KCl, 1mm DTT,0 .02 %N aN 3 and 15 %o f D 2 O. The 2D 1 H, 15 N-HSQC-TROSY spectra (in the presence or absence of peptide) were recorded with 32 scans and acquisition times of 200 ms for 1 Ha nd 100 ms for 15 N, at 30 8C. The spectra were analyzed in CCP NMR [25] and the chemical shift perturbation (CSP) were calculated according to the following equation: ,w here DH is the change in proton chemical shift, DN is the change in nitrogen chemical shift and 0.14 is as caling factor required to account for the difference in the range of amide proton and amide nitrogen chemical shifts. [26] A CSP was considered when it was greater than " x þ 2s.T he backbone assignment of VBC has been made available as by Dr.M ark Bycroft (Laboratory of Molecular Biology,M RC, Cambridge, UK) and shared as agift.
Supporting Information:T he raw BLI data are provided in the Supporting Information.